Generally, speech continues to gain importance in human-machine interaction. This may apply to multiple aspects of life, including, for example, voice-activated commands or phone conversations in a vehicle. Some vehicles may be equipped with multiple microphones in various locations in the car to allow multiple passengers to take advantage of voice-activated features. For instance, the driver may have a dedicated microphone, the passenger may have a dedicated microphone, etc. In this non-limiting example, it may be beneficial to determine which passenger is speaking, which may be difficult given the acoustic environment, close proximity of passengers, etc. However, current processes of Speaker Activity Detection (SAD) in multi-microphone systems may not be as robust, accurate, or as efficient as needed to satisfy the expectations of the end users.